The burdensome logistics of data linkage in Australia – the example of a national registry for congenital heart disease

Introduction

Data linkage is a valuable research tool, which involves the bringing together of information from various sources, creating a new, more comprehensive dataset.¹ Linked data is thus a very powerful resource in epidemiological research, and its benefits have been widely reported.^2–6 Data linkage allows for the study of disease groups across a population.^7–9 Matching records from the same individual across different data sources allows researchers to capture a ‘whole of life’ picture of a patient’s journey through a healthcare system.

The Australian Government reported on the substantial advantages of utilising linked administrative data in health research as part of the National Research Infrastructure Roadmap in 2006.¹⁰ This Roadmap also proposed the development of a coordinated national data linkage framework, resulting in the establishment of the Population Health Research Network (PHRN) in 2009.¹¹ The PHRN facilitates a network of the seven data linkage units (DLUs) representing the Australian states and territories, and a national DLU, the Australian Institute of Health and Welfare (AIHW), supporting researchers who wish to access linked population data. Multi-jurisdictional research projects are advised to apply via the PHRN.

Most administrative health service datasets involve hospital-based health care, such as admitted patient data. Much data on outpatient care can be obtained via linkage with the Medicare Benefits Schedule (MBS) and the Pharmaceutical Benefits Scheme (PBS) data. Linking of registry data to administrative datasets allows for analysis of the economic burden of disease, access to and patterns of care, and assessment of outcomes in a contextualised manner – taking into consideration the impact of health determinants such as ethnicity, remoteness, and socioeconomic status.¹²

In 2020, we were awarded an Australian Government Medical Research Future Fund grant, to analyse the burden of congenital heart disease (CHD) in Australia. The project involves the establishment of a National CHD Registry and includes funding for data linkage.¹³ The aim of our linkage project is to understand the burden of CHD from a healthcare provision perspective, and to provide more accurate monitoring of healthcare utilisation for CHD.

The study population consists of patients with a diagnosis of CHD derived from the 10 major CHD referral centres across Australia. The total cohort was defined as either Cohort 1, consisting of patients known to the clinics contributing to the Registry, or Cohort 2, patients not already known to the centres, but identified as having CHD by the DLUs through patient encounters associated with CHD diagnosis codes.

We have previously described the major bureaucratic inefficiencies in the establishment of a national disease registry in Australia.¹⁴ In this companion paper, we now examine the process of attempting to obtain data linkage for such projects.

The process of data linkage is currently complex and lengthy. For a national study, it can take >3 years to gain all required approvals, involving up to 8 DLUs, over 10 ethics committees, and numerous data custodians.¹⁵ This linkage project commenced at the end of 2021, with linkage unlikely to commence until early 2024. Although linked health data is an invaluable asset for researchers, the process of obtaining the necessary approvals is a major burden to applicant resources. This paper provides insight into the process of undertaking multi-jurisdictional data linkage in Australia.

Methods

Data collection

We reviewed the processes, required documentation, and applications required to conduct multi-jurisdictional data linkage across Australia.

Applying for multi-jurisdictional data linkage was a stepwise process (Table 1, Fig. 1). The first stage of the multi-jurisdictional linkage application process commenced with submission of an expression of interest (EOI) to the PHRN, via their online system. Once the DLUs are satisfied that the project is feasible, the project can proceed to jurisdiction-specific applications. Approval must be granted from each DLU linking the dataset, the data custodian governing the database, and the relevant ethics committee, and any other required bodies before linkage can commence.

Table 1.An overview of the data linkage application phases undertaken in the Congenital Heart Alliance of Australia and New Zealand Registry Data Linkage Project.

Consultation	Submission of an expression of interest via the PHRN’s online application system. Outline of the linkage project including aims, methodology, data flow, funding, any conflict of interests, consumer consultation, the research team, cohort selection, datasets of interest, data management and security, privacy and consent details, and an analysis plan Documents included: AIHW Technical Assessment, Registry study protocol, CHAANZ data dictionary, data linkage flowchart, and various jurisdiction-specific documents as outlined in Table 3.
Review	The PHRN meets with representatives of all involved data linkage units to discuss the project and review the expression of interest and associated documents. Feedback is given to the research team, and the expression of interest form is released for editing and resubmission. Submission of research team responses to each query raised by the PHRN and DLUs. Revision of expression of interest form and associated documents to address feedback.
Application	Preparation and submission of data linkage application/s and human research ethics application/s.
Approvals	Obtaining data custodian, ethics, and other necessary approvals (e.g. Public Health Act 2005 (Qld) approval for Queensland Health data, Public Interest Certificate for MBS/PBS data)
Data linkage	Data linkage units coordinate linkage of the approved data sets before delivery of data to the research team.

CHAANZ, Congenital Heart Alliance of Australia and New Zealand; DLU, Data Linkage Unit; PHRN, Population Health Research Network.

Fig. 1.

Flowchart of the application phase for all linkage units involved in the CHAANZ CHD Registry Data Linkage project. AIHW, Australian Institute of Health and Welfare; CVDL, Centre for Victorian Data Linkage; HREC, Human Research Ethics Committee; NT, Northern Territory; PHRN, Population Health Research Network; MBS, Medicare Benefits Schedule; PBS, Pharmaceutical Benefits Schedule; SA, South Australia; WA, Western Australia.

The AIHW National Hospital datasets are compiled from data previously supplied to the AIHW by state and territory health authorities. The Northern Territory (NT) does not participate in National Mutual Acceptance (NMA) for data linkage, and therefore NT data required a direct application to SA-NT Datalink and submission to a NT Human Research Ethics Committee (HREC). Also of importance to our study aims was the Victorian Cost Data Collection maintained by the Centre for Victorian Data Linkage. This dataset is not covered under the AIHW HREC, requiring submission to an additional NMA-participating accredited HREC.

The Western Australia (WA) DLU underwent major reform over 2022–2023, resulting in major delays and an extended closure of services. Due to these lengthy delays, Western Australian linkage was put ‘on hold’ until WA Research Data Service recommenced. The team re-engaged the WA linkage team in June 2023, although the WA DLU had a substantial backlog of requests, and our application has not yet been addressed at the time of writing and hence data pertaining to this process is not included in the results.

Results

Our EOI was first submitted to the PHRN on 1 December 2021. This commenced the consultation phase (Table 1). The project was authorised to commence jurisdictional ethics and governance applications on the 26 July 2022 (238 days from submission of the EOI), and the consultation and review phases were deemed complete on 13 March 2023 (468 days from submission of EOI).

Each DLU has its own processes within the application phase, as outlined in Fig. 1. The entire process took a total of 605 days from initial submission to final approval from all ethics committees.

Data linkage framework

Much of the consultation and review phases was spent developing and editing the data linkage framework, or data flow, of the linkage project (Fig. 2). This dataflow was complex, given that the project involved not only linkage of individuals in Cohort 1 whose identifiers were to be provided to DLUs, but also of the unknown Cohort 2. The data flow also had to be tailored to each state/territory DLU that was involved. The data flow went through several rounds of feedback with the DLUs. This feedback was seldom through the central PHRN feedback, with numerous requested changes being made via direct email or phone calls. Each time the data flow methodology changed, all documents involving the data flow had to be updated in the PHRN online system, as well as the direct communications being had with each DLU and ethics committees.

Fig. 2.

The data linkage framework required for data linkage for the CHAANZ CHD Registry linkage project. (a) Outline of the data flow for linkage with the AIHW and the Northern Territory DLU, (b) outline of the data flow for linkage with the AIHW and the Western Australian DLU, and (c) outline of the data flow for linkage with the AIHW and the Victorian DLU. AIHW, Australian Institute of Health and Welfare; CHAANZ, Congenital Heart Society of Australia and New Zealand; CHD, Congenital Heart Disease; CVDL, Centre for Victorian Data Linkage; FOD, Fact of Death; ID, identification; MBS, Medicare Benefit Scheme; MCD, Medicare Consumer Directory; NDI, National Death Index; NHMD, National Hospital Minimal Dataset; NNAPEDDC, National Non-Admitted Patient Emergency Department Data Collection; NT, Northern Territory; NTDL, Northern Territory Data Linkage; PBS, Pharmaceutical Benefit Scheme; PPN, Person Project Number; SURE, Secure Unified Research Environment; VAED, Victorian Admitted Episode Database; VCDC, Victorian Cost Data Collection; VEMD, Victorian Emergency Minimum Dataset.

Discussion

There are obvious important advantages of data linkage for national or state-based Registry projects, as this enables a complete picture of patient interactions with the healthcare system. This information is invaluable for patients, families, healthcare providers, systems planners, and governments. In this paper, we describe the continuing and substantial bureaucratic hurdles to data linkage in Australia, 15 years after the establishment of a centralised body which aims to streamline this process. These delays frustrate researchers but also, importantly, reduce Australia’s ability to generate meaningful data and reduce our international competitiveness in epidemiological research.

Without national data linkage capabilities, research cannot make use of cross-jurisdictional data linkage. Data linkage infrastructure has progressed since its inception,¹⁶ however, much duplication and complexity in the application process remains. Conducting research involving national data linkage in a timely manner is thus virtually impossible.

The difficulty in establishing cross-jurisdictional linkage problems lies, in part, within the complex multi-jurisdictional nature of Australia’s health system. Australia’s federal, state and territory, and local governments share responsibility for the health system, and consequently health data is not held centrally. Navigating the legal framework governing these systems, while achieving standardisation of data and linkage methodology across jurisdictions, is complicated.

Although all efforts were made to expedite our linkage applications, expected timelines were pushed back repeatedly. The PHRN online cross-jurisdictional application form was designed to enable a more centralised application process, but it is unclear if this system streamlines applications as intended. The central EOI process meant DLUs had some familiarity with the project once we progressed to the application phase, but nevertheless we had to undertake the same processes outlined by each individual DLU for single jurisdictional linkage (Fig. 1), not circumventing any steps after the lengthy consultation and review period.

The application phase involved substantial duplication of effort. Each time the data flow or application documentation was amended during consultation with individual DLUs, these changes had to be uploaded to the PHRN online form. This also occurred in the other direction, with changes in the PHRN form needing to be relayed back to DLUs. Multiple places to update versions led to confusion amongst the team and DLUs about document versions and changes to data flow.

Another limitation of this centralised process is that the progression of the consultation and review process is limited by the availability of all parties to meet and discuss the project. Feedback was not released to the research team until all DLUs had provided their feedback; this was thus limited by the slowest party. Significant delays were experienced due to the WA DLU undergoing reform, with the WA portion of the linkage eventually being placed on indefinite hold. This extended closure highlights the need for lasting resources to ensure the longevity of Australia’s linkage infrastructure. Unpredictable responsiveness added further delays to the project, with no expected response time outlined by most DLUs.

The research team was responsible for developing the data flow (Fig. 2), at times feeling like guesswork via many iterations of data flow documentation. It seemed counterintuitive that the data flow was not initially proposed to us by the DLUs conducting the linkage, who were the experts in data linkage, and were charging considerable dollar amounts for the linkage.

Other groups have also found that, although data linkage in Australia has overcome some challenges in developing the infrastructure for cross-jurisdictional data linkage, national data linkage projects are hindered by the complexity of the process.^15,17 A recent publication by Dutch researchers reported on obstacles similar to those faced in the Australian data linkage landscape.¹⁸ There is an urgent need for a truly streamlined application process with reduced duplication and greater consistency of procedures.^7,15,17,19

Improving Australia’s data linkage infrastructure requires greater homogeneity and simplification of the legislative framework governing health data. The Australian Government 2017 Productivity Commission inquiry report noted a risk-averse culture has led to ‘overly cautious interpretation of relevant legislation … and complex and lengthy approval processes’ (p. 55).²⁰ The 2017 report also noted a lack of national leadership and longwinded bureaucratic processes for data release.²⁰ While protecting patient privacy and ensuring ethical conduct is undoubtedly a critical element of research, the current approval process is a source of major inefficiency and limits the accessibility of data linkage.

Greater oversight from the PHRN could be considered. The PHRN provided excellent resources on navigating the complicated application phase, however, the requirements themselves were oftentimes unclear. One common application, such as the existing online PHRN EOI form, could be utilised by all DLUs, centralising the process.

The AIHW has been developing new data asset initiatives that should address some of the issues faced in accessing cross-jurisdictional health data, by integration of existing linked AIHW datasets to enable access in a more timely manner.²¹ This initiative is underway, with assets such as the National Integrated Health Services Information available for application, and the National Disability Data Asset being finalised.²²

Ultimately, the establishment of the data linkage project was a lengthy and intensive process. Access to linked data is invaluable to a National Registry, allowing for longitudinal analysis of population groups, however, the costly and time-consuming process of achieving the necessary approvals remains a challenge to both grant funding and timelines. Further streamlining, key accountabilities, and greater collaboration between jurisdictions is needed to ensure data linkage is feasible to researchers.

Data availability

Data sharing is not applicable to this article as no new data were created or analysed in this study.

Conflicts of interest

All authors declare that they have no conflicts of interest.

Declaration of funding

Funding has been provided by an Australian Department of Health grant through the Medical Research Future Fund, grant code is ARGCHDG0000028.